Shuffling Countermeasure against Power Side-Channel Attack for MLP with Software Implementation

In recent years, several attack methods for artificial intelligence (AI) have been reported. Therefore, the study of countermeasure against illegal attacks for AI is very important. This study proposes a new countermeasure method against power side-channel based model extraction attacks for AI. The proposed method improves the resistance against power side-channel based model extraction attacks by randomizing the calculation order with shuffling method. Experiments using an actual device indicated that the proposed countermeasure could improve the resistance against power side-channel based model extraction attacks.